Meta-analysis of T-cell receptor:peptide:MHC (TCR:pMHC) complex structural data available in PDB

DISCLAIMER You are free to use the processed and annotated TCR:pMHC tables in this repo for your research. However, the authors request priority in publishing a meta-analysis of the dataset aggregated and stored here. This message will be removed once this study is officially published. Thank you for understanding.

This repository contains a set of scripts that automate downloading, pre-processing and annotation of TCR:pMHC structural data with the following functionality:

scripts in annotation/src/ folder

• Extracts PDB entry metadata and verifies that a given PDB entry represents a valid TCR:pMHC complex
• Annotates each molecule in the complex: determines MHC class and allele, peptide molecule and TCR chains
• Performs V/D/J mapping for TCR chains, partitions TCR chain into CDR/FR regions

scripts in structure/src/ folder

• Computes pairwise amino acid distances and point energies for TCR and peptide residues

Running the pipeline

To run the pipeline execute the run.sh script. It will proceed with the list of PDB ids from result/extended_pdb_ids.txt. Some parts of the script are rather time-consuming, especially structural data annotation (downloading PDB files and running GROMACS). The results will be stored in result/ folder:

• final.annotations.txt contains the list of PDB entries that passed filtering and their annotation
• structure.txt or structure.txt.gz contains annotated data on the amino acid level, with pairwise residue distances and interaction energies for TCR:antigen pairs.
• structure.mhc.txt or structure.mhc.txt.gz contains annotated data on the amino acid level, with pairwise residue distances and interaction energies for TCR:MHC pairs.

Meta-analysis of the resulting dataset is stored in the analysis/ folder.

Pre-requisites

The pipeline is written in Groovy and Python (written in 3.5 but should run under 2.7) and requires both to run.

Three third-party software tools that are required:

• Anaconda or Miniconda with installed pandas and BioPython (for both condas) packages.
• Python packages openmm and pdbfixer, available only in Anaconda.
• BLAST. Ensure that blastp and makeblastdb are in your $PATH. We highly recommend you to use homebrew for OSX / Linux for BLAST and IgBLAST installations.
• IgBlast, strictly the 1.4.0 version. Ensure that igblastp is in your $PATH.
• GROMACS for computing interaction energies. Ensure that gmx is in your $PATH.